Dr. Katleen VRANCKX

Bioinformatics product specialist @ bioMérieux

Biography

Katleen VRANCKX joined the Belgium-based Applied Maths team in 2012 after obtaining a PhD in veterinary sciences at Ghent University. As a microbiologist and bioinformatics product specialist, Katleen has worked on improving BIONUMERICS based on customer insights and supporting customers to analyze their data with the highest efficiency. She was closely involved in the development of European foodborne disease surveillance networks, starting based on PFGE and currently transitioning to whole genome sequencing.

After 8 years of experience in analysis of whole genome data and MALDI, technical customer support and scientific communication Katleen joined the bioMérieux industry unit to apply her knowledge of public health strategies to improve food safety in the processing environment. 

Improve the typing possibilities of MALDI-TOF with advanced analytical methods

Matrix Assisted Laser Desorption-Ionization Time of Flight Mass Spectrometry (MALDI ToF MS, here “MALDI”) is a technique most of us know for the identification of bacterial species. For unrelated species, this is an easy analysis, even by eye the spectra look completely different. Most analysis methods have no problem differentiating related species. Many have wondered if we can go beyond the species and also differentiate very closely related species and even strains. The answer is not straightforward and in most cases it is a ‘maybe’.

Clustering methods typically do not cut it for this purpose, as these will cluster the strains using the complete peak set while our ‘strain’ signal is often just found in a few peaks. Instead, we will use non-hierarchical methods implemented in BIONUMERICS for the analysis, starting with a linear discriminant analysis on a set of reference strains for which we know the type. This will provide an answer whether a specific ‘strain’-signal is present or not. If this signal is indeed present, we can continue to more advanced machine learning methods. Our reference set can be used to train the method to predict the strain on unknown profiles. A dataset of different isolates from several species in the Mycobacterium tuberculosis complex, notoriously difficult to differentiate with MALDI, is used to demonstrate this approach. The majority of isolates were classified to the correct species, with a few misclassifications in M. africanum and M. bovis, mainly due to the underrepresentation of these species in our reference set.

In conclusion, this approach will provide us based on a reference set, whether we can distinguish types and if so, with how much confidence.

Partager cette page :